All movements are controlled by the graded activation of different muscles. Muscle activation is controlled by the activation of motoneurons in the brainstem and spinal cord. Each motoneuron drives the muscle fibers it innervates in a one-to-one fashion. Muscle fiber action potentials are relatively easy to record in human subjects and it has long been appreciated that the firing patterns of motor units provide a unique window on the properties of motoneurons and their inputs. The goal of this proposal is to deduce the pattern of inputs to motoneurons based on an analysis of their output. The success of this reverse engineering approach depends upon two factors: (1) the existence of accurate models of the input-output properties of a population of motoneurons, and (2) the ability to extract the maximum amount of information from motoneuron output, based upon recording the activity of multiple, simultaneously-active motor units. We have recently developed a set of motoneuron models that can replicate several key features of the input-output properties of motoneurons with different intrinsic excitability. When these models are driven with different patterns of excitatory and inhibitory inputs they reproduce a number of the features of the discharge of human motor units during voluntary contractions. Previously, recording the behavior of multiple units required combining recordings taken across multiple sessions and subjects. New developments in recording and analysis techniques now make it possible to record from 10 or more motor units in each trial. This provides a rich data set that can be used to constrain the set of input patterns that give rise to a given pattern of output. The overall hypothesis of this proposal is that the firing patterns of a population of motoneurons are determined by the patterns of their synaptic inputs and the level of neuromodulatory drive, making it possible to deduce input patterns from the firing patterns of multiple motor units. We will test this hypothesi by comparing our estimates of input based on the reverse engineering approach with direct voltage-clamp measurements of the inputs in the same experimental preparation. There are three specific aims (1) to improve techniques for recording the simultaneous activity of multiple motor units, (2) to validate our reverse engineering approach by comparing synaptic input patterns that are predicted from recordings of motor unit discharge to those that are directly recorded from motoneurons, and (3) to determine the role of synaptic inhibition in shaping motor unit discharge patterns, and the ability of our reverse engineering approach to detect different patterns of inhibition. Once successfully developed in our animal preparation, the reverse engineering methods can be adapted for use in human subjects, allowing maximal use of the rich information available in motor unit firing patterns for understanding the structure of motor commands in humans in both normal and pathological states.